CLIPP
Christiani Lehmanni inedita, publicanda, publicata
titulus
Language description and general comparative
grammar
huius textus situs retis mundialis
http://www.christianlehmann.eu/publ/lg_descr.pdf
dies manuscripti postremum modificati
14.12.2012
occasio orationis habitae
11th International Congress of Linguists, 13.08.1987, Berlin
volumen publicationem continens
Graustein, Gottfried & Leitner, Gerhard (eds.), Reference
grammars and modern linguistic theory. Tübingen:
Niemeyer (Linguistische Arbeiten, 226)
annus publicationis
1989
paginae
133-162
Language description
and general comparative grammar
Christian Lehmann
Abstract
The following postulates are formulated with respect to a scientific description of
a language:
1. It is usable for whatever linguistic purpose.
2. It comprises an account of the linguistic system, a lexicon, a text corpus and a
statement of the historical situation of the language.
3. The description of the language system accounts not only for core, but also for
peripheral subsystems.
4. The linguistic system and the lexicon are presented both in a synthetic and in
an analytic form.
5. The description brings out the dynamic character of the language.
These postulates can be complied with if the description of the language system
instantiates a general comparative grammar. This in itself obeys the postulates.
Specific proposals for the implementation of a general comparative grammar, esp.
with respect to postulate 4, are made.
Christian Lehmann
2
1. Introduction
The form of a model of the syntactic structure of a language has been the subject of extensive
discussion in the confines of certain theories of grammar. The question, however, of how a
comprehensive description of a language is to be devised has, to my knowledge, received
surprisingly little attention from linguists, either in formal terms or at a general level.1 This
is the basic problem of the theory of linguistic description (whose central component is the
theory of grammar).
However, the question is not just a theoretical one. Answers to it have to be inspired
empirically by actual language descriptions. In fact, the general problem of a language
description is at least as much one of general comparative linguistics as it is one of the
theory of linguistic description. In what follows, the position will be advocated that the
elaboration of a model of linguistic description is bound up with the elaboration of a general
comparative grammar.
The presentation is organized as follows. In §2, a set of requirements to be met by any
language description is derived from a definition of `language description' and a thesis on its
composition. These are put forward in the form of postulates (numbered consecutively with
a capital P). In §3, the scope is broadened to include further postulates and theses (with a
capital T) which connected with the requirement that a language description should be
comparable with those of other languages and therefore be based on a general comparative
grammar. This is discussed partly on the basis of a critical evaluation of the Lingua
Descriptive Studies Questionnaire. In §4, the organization of the general comparative and the
language-specific grammar according to the synthetic and analytic viewpoints is explained.
2. Demands on a language description
In §2, I will formulate and argue for a set of postulates and theses which circumscribe the
contents and form of a language description.
2.1. Aim
Definition:
A language description is an encyclopedic description of a language the
contents and form of which are defined with respect to linguistics as a science.2
The linguistic description here envisaged scientifically describes one natural language, the
1
Among the exceptions that I am aware of are Gabelentz 1901, Zweites Buch, Seiler 1969, Lehmann 1980
and Mosel 1987.
2
Science is, of course, to be taken in the sense of `Wissenschaft', not in the sense of `Naturwissenschaft'.
Language description and general comparative grammar
3
object language, in terms of another natural language, the background language (also
called metalanguage).3 It is one kind of language description.
Other kinds are, e.g., a pedagogic description or a general purpose description for an
encyclopedia. Disregarding these for the rest of the discussion, I will follow the above
definition and henceforth call the linguistic description of a language simply a language
description. It must be usable by any linguist for whatever his specific interests may be. This
implies, among other things,
a. that no knowledge of the language or any aspect of it (e.g. its writing system) is
presupposed;
b. that the general state of the art in linguistics is presupposed, so that a layman is not
expected to be able to use the description;
c. that the description will not be framed in terms of some formal model, but in terms
intelligible to any linguist, and that practical considerations of usability shape it to some
extent;
d. that didactic considerations play a minor role in the presentation, since the description must
be usable as a manual.4
2.2. Composition
P1. A language description is a comprehensive presentation of a language under all its
aspects.
The term `comprehensive presentation' is intended to include not only a scientific analysis
or its result, a model, but also a documentation of representative specimens of its subject
matter. This leads to the following thesis:
T1. A language description consists of four parts:
1. description of the language system,
2. lexicon,
3. text corpus,
4. description of the historical situation of the language.
Let us now elaborate on these four parts in turn.
3
I call it background language rather than metalanguage because it is the language in terms of which we
understand the object language. It is also used in the dictionary (cf. §2.2 ad 2), where it would not normally
be said to have the status of a metalanguage in the scientific sense.
4
Most grammars do indeed present information by stepwise building up complex structures from simpler
ones. The Lingua Descriptive Studies Questionnaire, to be discussed in §3.2, and the grammars based on it
represent the most blatant and remarkable violation of this tendency, by starting with sentence types and
subordination.
4
Christian Lehmann
Ad 1. The description of the language system comprises
a. the phonology with its interfaces to phonetics and orthography,
b. the grammar stricto sensu, i.e. morphology and syntax,
c. the semantics with its interfaces to pragmatics and stylistics.5
The interfaces are described to the degree that they are systematic.
Given that, according to the definition in §2.1, a language description is constituted inside the
science of linguistics, signs of the language should be given in phonological representation.
Here, however, the practical considerations also mentioned there come into play. Provided
that the orthographic representation preserves the identity of the sign, it is to be preferred. On
the other hand, if the orthography is Non-Roman, a transliteration is the best solution for an
alphabetic script, a phonological representation, for a non-alphabetic script.
Ad 2. Again given the definition in §2.1, the lexicon should take the form of a lexicological
description. However, on the one hand, the lexicon by definition contains what is
non-systematic about the signs of the language; insofar the list form is adequate to it. And on
the other, for ease of reference, the lexicological description would have to be supplemented
by an index, anyway. Practical considerations therefore speak in favor of representing the
lexicon in the form of a "bilingual" dictionary, which associates the object language with the
background language. However, such a bilingual arrangement also has some properly
linguistic motivation, for which see §3.3.
Each entry contains information of the following kinds (cf. the lists of T1 and ad 1 above):
a. phonological, incl. phonetic and orthographic,
b. grammatical (i.e. morphological and syntactic),
c. semantic, incl. pragmatic and stylistic,
d. historical, incl. all the rubrics mentioned ad 4.
e. examples, with reference to the text collection.
Again mainly for practical reasons, the lemma is given in orthographic representation, which
entails that phonological and phonetic information (item a) will be specified only to the extent
that it is idiosyncratic with respect to the orthography. Analogous considerations apply to the
pragmatic and stylistic information.
Ad 3. There are various reasons why a language description should include a text corpus.
5
Pragmatics is here taken to include all kinds of non-linguistic knowledge which determines the use and
interpretation of utterances, in particular socio-cultural conventions as they have been described variously
in sociolinguistics, the ethnography of communication or the theory of actions. Thus, pragmatics is not part
of the linguistic system (or its description). It is distinct from functional sentence perspective (which
pragmatics reduces to in some terminologies), which is accommodated in the present model in §4.1.
Language description and general comparative grammar
5
First, the texts confirm or falsify the statements of the description. Second, our techniques of
analysis at levels above the sentence are presently so little advanced that it is safe to
supplement the efforts in that area by simple ostension. Third, while grammar and lexicon
present the system, the texts represent the norm, or various norms, in the sense of Coseriu
1952. Especially for the latter reason, it is essential that the texts illustrate different forms of
communication (text sorts); cf. ad 4, b. They should not all be narratives; there should be
speeches, instructions, rituals, plays, jokes and, above all, dialogues.
The constitutive parts of each text are the following:
a.
b.
c.
d.
orthographic representation,
interlinear glossing, providing morphological and possibly syntactic information,
translation into the background language,
commentary.
These kinds of information bear some correspondence to the items ad 2. As in the other parts
of the description, orthographic representation is chosen for practical reasons. Also, it is
normally closer to the morphophonemics than a purely phonological representation and thus
facilitates the interlinear glossing. The latter should be done according to the guidelines set
out in Lehmann 1983. The translation itself can then be quite idiomatic. The commentary
accounts for any aspects of the texts whose understanding is not derivable from the other
three parts of the description.
Ad 4. A language could be regarded as a system established over an inventory of signs from
which texts may be constructed. Insofar, it would be accounted for by the first three parts of
the description. However, an adequate notion of a natural language is more comprehensive
than that. A language is bound up with the life and culture of its speech community; it is a
historical phenomenon. As such, it is not adequately grasped by a purely structural
description. The global historical situation of the language has to be described as well. This
account may be subdivided into the inner and outer situation as follows:
a. inner: dialectal and sociolectal varieties, history, genetic affiliation of the language;
b. outer: ethnographic aspects of the speech community itself, role of the language in
the society (e.g. in a multilingual situation), conventional forms of communication,
literature.
The description of the language system and the lexicon together form the core description.
All the components of the language description are tightly linked to each other. It is general
practice that the description of the language system (especially, the grammar) and the lexicon
complement and therefore constantly refer to each other. The same is true of the other parts.
In particular, the core description refers to the text corpus for further examples. The lexicon
also refers to the statement of the historical situation for points of fact. The texts of the corpus
Christian Lehmann
6
illustrate, and the commentaries accompanying it refer to, the statement of the historical
situation. F1 shows the composition of the language description schematically.
F1. Composition of a language description
Core description
1. Language system
[global subdivision]
2. Lexicon
[subdivision of entry]
stylistics
3. Text corpus
[representation levels]
pragmatics
4. Historical situation
[global subdivision]
outer situation
semantics
translation
grammar
syntax morphology
interlinear glossing
inner situation
phonology
phonetics
orthography
orthographic representation
2.3. Completeness
P2. The description of the language system is complete in its account of the - central and
peripheral - subsystems.
Every linguistic category and subsystem has the status of a prototype. This means that it has
a center and a periphery (cf. Dane÷s 1966). Even the linguistic system as a whole has
peripheral subsystems. What these are should be subject to empirical research. However,
there have been inveterate preconceptions in linguistics which regard certain subsystems such
as the phoneme system, parts of speech, morphological categories, syntactic relations and
clause structure as central.6 Numerous grammars, among them even some generally regarded
as good specimina of their kind, limit themselves to an account of these. One often misses
word formation, very frequently complex sentences, almost always particles (modal particles,
interjections, ideophones etc.). I know of no linguistic theory which proves that these areas
are peripheral to the linguistic system, although they may be.
Given that a language is not systematic throughout and, even as a system, is open in every
direction, it cannot be described exhaustively. However, the theory of linguistic description
has to guarantee that all the subsystems of a language are accounted for to the degree of their
6
Some of these preconceptions may be based on theories which confuse centrality with regularity.
Obviously, the regular parts of a language are more easily described than the irregular ones.
Language description and general comparative grammar
7
relevance to its functioning.
In the last years, general comparative linguistics has drawn our attention to phenomena which
used to be neglected in earlier language descriptions. Consequently, linguistic comparison
also serves as a heuristic tool to guarantee the completeness of the description.
2.4. Hermeneutics and dynamicity
P3. A language description renders the object language intelligible.
Language is a human activity whose immediate goal it is to make sense. This goal is common
to speaker and hearer. The linguist has to respond to it. His description has to bring out the
sense that is hidden in linguistic structure. To this extent, his task is an interpretative or
hermeneutic one.
In the present framework, the hermeneutic quality of a description takes the place of what is
termed explanation in other frameworks. Instead of trying to explain an instance of language
by positing laws for it, we should try to understand it or, rather, to show how it is to be
understood (cf. Lehmann 1987, §4.2 for some examples).
To understand a human act (e.g., an utterance) means to know its goal and the conditions
under which one might do or have done that act oneself (if one had the ability). Analogously,
to understand an activity (in particular, a culture-bound activity such as a language) means
to know the circumstances under which this activity has to fulfill its goal, so that one might
engage in it (if one had the ability).
Knowledge of the conditions of an act includes knowledge of the available alternatives. A
hermeneutic linguistic description therefore accounts for variation. Knowledge of the
circumstances of a social activity includes knowledge of its direction in time, of the sense in
which it is currently developing. Consider, for illustration, the projection of a movie. It is
normally impossible to fully understand a still picture if one has no knowledge of the
segments immediately preceding and, perhaps, following it. An "ideal" synchronic linguistic
description in the sense of a pure momentary cross-section therefore is not a hermeneutic
description. This leads to the following postulate:
P4. A language description represents the operational and evolutive dynamism of the
language.
As an activity, language is dynamic. For the synchrony, this entails that a language is not
reduced to an inventory of categories and relations,7 but that it consists equally of operations
7
This reduction is one of the few things in Langacker's (1987) cognitive grammar that are incompatible
with the present framework.
8
Christian Lehmann
which associate functions with structures, which select items from their paradigmatic class
and combine them with their syntagmatic context. For the diachrony, the dynamicity of
language entails that at any given moment some components (grammatical concepts,
expression devices and operational strategies) are declining, others gaining ground. This is
the presence of diachrony in synchrony or what Sapir called the `drift' of a language. Cf. also
Coseriu's (1987:46f) `principle of dynamic description'.
Linguistic description accounts for this by ordering grammatical concepts and expression
devices on continua. Grammaticalization scales play a prominent role here. Since they
incorporate both expression and content of the language sign, they can serve as an
organizational principle both in the synthetic and in the analytic part of the description. A
description which accounts for the dynamicity of language acquires diachronic depth without
mingling different historical stages. It represents the diachrony in the synchrony.
3. General comparative and language-specific grammar
There is a mutual dependency between general comparative linguistics and descriptive
linguistics, which proves sometimes painful, more often fruitful. On the one hand, progress
in general comparative linguistics depends on the availability of good language descriptions.
On the other hand, progress in descriptive linguistics depends on the availability of good
theories of linguistic description, the central part of which is expected from general
comparative linguistics.
The kind of relationship that I assume between general comparative and specific grammar is
formulated in P5, which is meant as a postulate to be followed by the descriptive linguist.
P5. Describe your language in such a way that the maxim of your description could serve,
at the same time, as the principle of a general comparative grammar - and, thus, as the
maxim of the description of any other language.
P5 is the categorical imperative of language description. It naturally leads to the following
thesis:
T2. The description of a specific language is a concreticization of general comparative
grammar.
The following subsections will establish T2 and show how P5 can be complied with.
3.1. General comparative grammar
Most linguists, I presume, will find it desirable that a language description conform to the
definition and postulates put forward in §§2.1 - 2.3 and, possibly, 2.4. Everybody, not only
the general comparative linguist, wants a language description to be comprehensive. There
Language description and general comparative grammar
9
are very few, if any, language descriptions which are complete in the sense explained in
§2.2f. Most of the time one has to gather the information wanted from different and
independent sources. In part, the completeness of a language description depends on the
resources available to the analyst. For another part, however, it depends on the availability
of information on the contents and form of a comprehensive language description and how
it is to be done. As long as such information is not available, no linguist can blame any other
for the unsatisfactory state of affairs.
The general comparative linguist has another demand to pose upon a language description
which we may formulate in the following postulate:
P6. A language description is comparable with the descriptions of other languages.
If every analyst organizes his language description according to a personal scheme, his
description may be as comprehensive as may be; but it will be difficult to compare
information drawn from it with information drawn from the description of another language.
Again, as long as no generally applicable schema is available, the descriptive linguist can
hardly do any better.
Requirements both of comprehensiveness and of commensurability of language descriptions
thus converge in the call for a general schema on which the description of any language can
be based. This is what used to be called a general grammar. In the heyday of rationalism,
a general grammar was conceived as a deductive enterprise. As such, it would have to be
grounded in a general theory of language. However, to the degree that such a deductive basis
was not in fact available, the general grammars of those times were largely based on
principles of logic and of the normative grammar of certain classic Indo-European languages.
With the advent of linguistics as an empirical science, this pastime was abandoned. We
witness its partial resurrection in our day in the form of so-called universal grammar.
In the last decades, much comparison of languages has been done on an empirical basis. The
Dobbs Ferry Conference on universals of language (cf. Greenberg (ed.) 1963) has stimulated
large amounts of research which strive for contributions to general grammar by accumulating
inductive generalizations over the languages compared. This approach, in turn, has been
criticized for being atheoretical, for putting out sets of largely isolated observations and
hypotheses whose relevance to a general theory of language is not clear.
From this account, it becomes obvious that the general grammar that we want will have to
combine a deductive and an inductive approach. This is why I call it general comparative
grammar (GCG). It does not yet exist, but important contributions towards it have forthcome
both from the deductive and from the inductive approaches.
Each of the four parts of language description presented in §2.2 should be the subject matter
of general comparative linguistics. Thus, there should be a general comparative discipline
10
Christian Lehmann
occupying itself with text corpora;8 and there should be one dedicated to the historical
contexts of languages. These will be neglected here. A general comparative model of
language structure will comprise the core description. GCG stricto sensu refers only to the
central portion of this model, the grammar as opposed to semantics, phonology and lexicon.
In what follows, I will restrict my attention to GCG in this narrow sense (for contributions
to general comparative lexicology, see Talmy 1985 and Lehmann 1988[P]).
3.2. The Lingua Descriptive Studies Questionnaire
Lingua Descriptive Studies (1979-1982, since 1984 Croom Helm Descriptive Grammars) is
a series of grammars of different, mostly exotic, languages which are all organized according
to the same schema. The schema was published in 1977 by the editors of the series, B.
Comrie and N. Smith, in the form of a questionnaire. The Lingua Descriptive Studies
Questionnaire (LDSQ) does not pretend to be a GCG. It is just meant to be a systematic
catalogue of questions the answering of which would secure the completeness and
comparability of the descriptions. According to its introduction (p. 5), "it is important that the
general framework be sufficiently flexible to enable any arbitrary language to be described
within this framework". However, to the degree that the LDSQ approaches this aim, it does
represent an important contribution towards GCG - incidentally one reflecting predominantly
the empirical-inductive approach. I will briefly discuss its overall organization, so that the
requirements to be fulfilled by a GCG become more apparent.9
The main subdivision of the LDSQ is in
1. Syntax
2. Morphology
3. Phonology
4. Ideophones and interjections
5. Lexicon.
Part 4 just asks for the ideophones and interjections of the language and is not subdivided
further. The lexical part asks for the items of a couple of semantic fields and for 207 items
of basic vocabulary. Thus, the bulk of the conceptual work has been invested into parts 1 3.
The main section headings of part 1 are:
8
Apparently, next to nothing is known about how a representative text collection of a language is
composed.
9
I am aware of one discussion of the LDSQ in the literature, namely Uhlenbeck 1980. This is essentially
concerned with the objection that the questionnaire should not presuppose the universal existence of
conceptual categories to be variously expressed in languages.
Language description and general comparative grammar
11
1. General questions [i.e. sentence types and subordination]
2. Structural questions [i.e. clause and phrase structure]
3. Coordination
4. Negation
5. Anaphora
6. Reflexives
7. Reciprocals
8. Comparison
9. Equatives
10. Possession
11. Emphasis
12. Topic
13. Heavy shift
14. Other movement processes
15. Minor sentence types
16. Operational definitions of word-classes.
The main section headings of part 2, with headings of section 2.1 shown for clarity's sake,
are:
1. Inflection
1.1. Noun inflection
1.2. Pronouns
1.3. Verb morphology
1.4. Adjectives
1.5. Prepositions/postpositions
1.6. Numerals/quantifiers
1.7. Adverbs
1.8. Clitics
2. Derivational morphology.
The main section headings of part 3 are:
1. Phonological units (segmental)
2. Phonotactics
3. Suprasegmentals
4. Morphophonology (segmental)
5. Morphophonology (suprasegmental).
So much should suffice to give an idea of the general disposition. According to the
introduction (p. 8), "the general direction of description within the questionnaire is from
function to form." However, this principle is not at all observed consistently in the LDSQ. Its
12
Christian Lehmann
first violation lies in the subdivision between syntax and morphology. Comrie & Smith, just
as most of us, draw the boundary between these two with respect to the grammatical level of
the word. Now this is clearly a structural criterion. Consequently, all the function-form
oriented questions posed inside the morphology chapter presuppose that any language to be
described will fulfill the corresponding functions at a certain structural level. For instance,
the question for definiteness/indefiniteness marking is asked in the section on noun inflection.
What then shall we say of languages such as English, which express this distinction by
separate words (after all, articles are not part of noun inflection) or, worse, of languages such
as Russian which express a similar distinction by word order?
The same problem repeats itself at the lower hierarchical levels of the LDSQ. To resume the
last example, there are languages (such as Hungarian) which mark definiteness by verb
agreement. For their description, the question as to the expression of definiteness would
properly belong into the section on verb inflection. Similarly, the question as to the
expression of the syntactic functions of noun phrases is also posed in the noun inflection
section, although it might as well be asked in the verb inflection section. The curious
consequence of this organization of the LDSQ is that a large part of the verbal morphology
of several languages described in the Lingua Descriptive Studies series (e.g. of Abkhaz, cf.
Hewitt 1979) is presented in the section on noun inflection.
Obviously, such problems could be solved within the LDSQ framework. It is not necessary
that a question concerning definiteness presuppose noun inflection; it might be asked
appropriately in a chapter on deixis and reference. Similarly, syntactic functions of noun
phrases will naturally emerge from a discussion of the functional domain of participation, i.e.
the articulation of an event in terms of core and participants. A purely function-form oriented
GCG would be feasible. What, then, if the LDSQ were organized consistently in such a way?
A purely function-form oriented approach inevitably leads to the consequence that the
different uses of a polysemous or multifunctional form are dispersed over the various
chapters. For instance, in a description of English, the verb to be would have to be treated in
ch. 1.2 (clause structure, namely copular sentences), 1.10 (possession), 1.11 (emphasis,
namely clefting), 2.1.3 (verb morphology). Comrie & Smith acknowledge this problem and
ask contributors to counteract it, chiefly by extensive cross-referencing.
Two objections may be raised against this procedure. First, whatever the analyst can do
within the predominantly function-based framework in order to alleviate the above problem,
will remain patchwork, since it is not systematically provided for in the conception. One can
understand a language linguistically only if one has a coherent picture of its functioning in
its own terms. This much, at least, appears to be acceptable of the justification for the purely
form-based American structuralist grammars.
Language description and general comparative grammar
13
Second, there is no reason why a grammar - a GCG or a specific grammar - should be
preferably function-based rather than form-based. Comrie & Smith assume that general
comparative linguists will use Lingua Descriptive Studies essentially with such questions in
mind as `How is the concept of definiteness, or of possession, or whatever, expressed in this
language?' This is doubtless a frequent and important kind of question asked by such
linguists. However, there are also questions such as `What does verbal prefixation, or vowel
alternation, or front shift of constituents, express in this language?' The grammars of this
series answer such questions only insofar as they, or the LDSQ, are really not purely
function-based and, thus, inconsistent.
This discussion of a well-known and meritorious contribution to GCG leads us to P7.
3.3. Analytic and synthetic viewpoints
P7. In a language description, those items have to be treated together which are similar in
the object language.
P7 is a basic requirement to be met by any language description. It is based both on
grammar-theoretical reasons of adequacy and on practical reasons of usability. Certain
directions of American structuralism have inferred from P7 that the description has to follow
exclusively a principle which results from the structure of the language in question. This
conclusion proved to be undesirable, as such a presentation renders the description unusable
for non-specialists in the language. Moreover, it does not really follow from P7.
A weaker thesis, however, does follow from it, namely that the peculiar structure of the object
language should be brought out by the description. This follows also from P3. Now this thesis
immediately seems to conflict with P5 and T2, since the latter amount to the requirement that
every language should be described according to one general schema. §§3.4 and 4 will be
devoted to the resolution of what has proved to be a basic dilemma in descriptive linguistics.
Here we may observe that a very similar dilemma arises already inside the particular language
description which tries to obey P7, quite regardless of any requirements of general grammar.
All elements of a language which relate both to expression and to content may be similar in
either of two respects: they may be functionally similar or structurally similar.10 The
association of function with structure in any language is partly motivated, partly arbitrary. To
the degree that it is motivated, functional similarity correlates with structural similarity. To
the degree that it is arbitrary, functionally similar elements are structurally dissimilar, and vice
10
Although the most general terms for the two sides of the language sign are expression (significans) and
content/meaning (significatum), it is customary to speak, with respect to expression and content within
grammar, of form/structure and function, instead.
14
Christian Lehmann
versa. If this is accepted, P7 entails the following thesis:
T3. The core description (cf. F1) is carried out according to two complementary viewpoints,
the analytic and the synthetic.
In one part, the principle of the disposition of the material is a formal-structural one. One
starts from the structures, interprets these and thus arrives at the functions. The corresponding
part of the lexicon leads from the object language to the background language. This
corresponds to the viewpoint of the hearer or of a user who is confronted with a text in the
language and wants to understand it. This is the form-function oriented or analytic part of
the description.
In the other part, the principle of disposition is a functional-semantic one. One starts from the
functions, looks for their realization and thus arrives at the structures. The corresponding part
of the lexicon leads from the background language to the object language. This corresponds
to the viewpoint of the speaker or of a user who wonders how a given function is fulfilled in
this language. This is the function-form oriented or synthetic part of the description.
Not only the lexicon and the grammar, but each of the three sections of the language system
as shown in §2.2, ad 1, are described according to the two complementary viewpoints. This
is shown in F2.11
11
This conception originates with Gabelentz 1901:84-125. A similar one is reflected in Sapir's (1921, ch.
IVf) distinction between grammatical processes and grammatical concepts, which itself goes back to F. Boas.
It may be useful to recall that trends in the recent history of linguistics partly differ in their preference for one
of the directions of description. Thus, early American structuralism worked "from phoneme to utterance",
while generative semantics worked from semantics to phonology. Cf. also the directionality debate of the late
sixties in generative grammar. Another pair of terms belonging into this context is `recognition grammar vs.
production grammar'. For further discussion, cf. Jespersen 1924:39-46, Lehmann 1980, Mosel 1987.
Language description and general comparative grammar
15
F2. Analytic and synthetic viewpoints
stylistics
pragmatics
semantics
)) grammatical functions ) )))))))))))) ))))) significata )))))
grammar
syntax
*
lexicon
morphology
inventory
*
)) grammatical structure ) )))))))))))) ))))) significantia )))))
phonology
<))))) synthetic )))))
))))) analytic )))))>
))))))))))))))))) semantic representation )))))))))))))))))
)))))))))))))) systematic phonetic representation )))))))))))))
phonetics
orthography
The analytic and the synthetic viewpoints are based on entirely distinct and independent
systematics, which will be discussed in detail in §4. Each of them could found a language
description by its own, which would then be one-sided in the ways discussed. A complete
language description consists of both the analytic and the synthetic systems. If the description
is published in book form, the twofold organization would manifest itself in two major parts.
One could, however, imagine an electronic implementation of the description, e.g. in a
knowledge representation system, which is one complex whole and where the two viewpoints
are implemented as two alternative paths of access to the same information.
3.4. General comparative grammar and the specific language system
A GCG is a maximum model12 of what may be found in natural human languages. A
maximum model is, of course, not an accumulation of features from diverse languages.
Instead, it embodies a systematicization of the observable cross-linguistic variation. R.
Jakobson's hierarchies of unilateral foundation constitute a clear case in point. The variants
are represented, but they are not just enumerated. Instead, the principle of the variation is, at
the same time, the principle of the disposition of the material in the GCG. Moreover, a certain
level of abstraction from language-specific detail is necessary. Trivially, two case systems say the English and the French ones - may be said to represent the same type, in spite of
obvious differences, and thus not be accounted for separately in the GCG.
A GCG is based upon a theory of language universals. A language universal is something
12
Cf. Dressler 1967 for this concept.
16
Christian Lehmann
which is an essential constituent of language and, therefore, represented in every language.
A theory of language universals is a central part of a theory of language. Structural properties
of languages have, in principle, the theoretical status of variants;13 as such, they are generally
not language universals. Consequently, while structural properties of language systems are
represented in a GCG, they are generally not represented in a theory of language universals.
Consequently, a GCG is clearly distinct from a theory of language universals. The expression
`universal grammar' occasionally found in contemporary general linguistics is a misnomer for
either of these two things.14
In §3.3, it was postulated (P7) that a language description must treat together what is similar
in the object language. In §3, it was maintained (T2) that a language description be based on
a general comparative grammar. Now a language is not just an eclectic accumulation of items
from the superset of all possible language properties; it is a system "où tout se tient".
Therefore, the question arises how the two propositions are reconcilable.
The problem whether a language description based on a non-language-specific grid can
represent the spirit of the individual language is a very real one. Experience with a dozen of
volumes of Lingua Descriptive Studies has shown that a general framework cannot possibly
foresee all the fanciful associations of functions with structures that appear in the various
languages.
However, we have to differentiate between polysemy and homonymy. At a cross-linguistic
level, the distinction between the two can be reformulated as a distinction between a
content-expression association recurring in unrelated languages and one occurring in only one
language. Take two German examples: a) The word selbst functions both as an identifier
(`self') and as an emphasizer (`even'). b) The suffix -er functions both as a plural marker on
declinable words and as a comparative marker on adjectives. An internal analysis of German
here clearly brings out the polysemy of case a, the homonymy of case b. Cross-linguistic
evidence confirms this. Identity of elements functioning in identification and emphasis recurs
in Portuguese (mesmo), Tamil (taan) and many other languages.15 Identity of plural and
comparative morphemes recurs only in languages genetically related to German.
Thus, one of the requirements to be put on a GCG will be that it be organized in such a way
as to bring out the connection between identification and emphasis. And, in fact, for most of
the cases, it will do this both in its synthetic and in its analytic portion. In the synthetic
system, identification of an entity with itself and emphatic underscoring of it will be treated,
in the functional dimension of reference, as two similar kinds of relationship of an entity with
the universe of discourse. In the analytic system, the two uses of the morpheme in question
13
This has been emphasized by H. Seiler since 1972.
14
I have argued this repeatedly, e.g., with special reference to a couple of established strands of language
universals research, in Lehmann 1982.
15
Cf. Moravcsik 1972 and Edmondson & Plank 1978 for data and explanations.
Language description and general comparative grammar
17
will be treated in the same or at least in adjacent chapters to the degree that the distributions
of the morpheme in the two cases are similar. German selbst, e.g., is in any case a sentence
level particle with order properties describable by quantifier floating.
The conclusion is that the framework of GCG will allow to bring out the spirit of the
function-form association of a specific language to the degree that the individual associations
are functionally or structurally motivated. It will not provide a common denominator for cases
of homonymy; but then there is none, in the first place.
4. Synthetic and analytic systems of GCG
The discussion in §3.2-4 was intended to support the thesis (T3) that a GCG, just like an
individual grammar, has to be organized according to the analytic and the synthetic
viewpoints. In the following sections, this proposal will be fleshed out.
4.1. The synthetic system
The synthetic part of the core description is organized independently of linguistic expression
structures and, instead, grounded in human cognition and semiosis. The set of concepts and
operations represented in human languages is structured in various ways. We can assume it
to be articulated, at the highest hierarchical level, in what may be called cognitive domains.16
Most of these are represented only in the lexica of languages. Examples are the domain
comprising the physical constitution of the cosmos, the life cycle or the ingredients and
operations of cooking. Some of them, however, reach over into the grammar; indeed, they
provide the semantic subject matter of human grammars. To these belong the domains of
possession, of spatial and temporal orientation and a limited set of others, to be reviewed
presently.
There are two ways in which one might misunderstand this conception. One would be to
assume that the lexicon, or lexical meanings, are "more universal" than the grammar, or
grammatical meanings. What is meant, instead, is that all universal conceptual domains find
their manifestation in the lexicon of every language, but some of them have no relevance for
the grammatical structuring of languages. Another misunderstanding would be to conclude
that the entire lexicon of every language is somehow contained in the set of universal
cognitive domains, or reducible to universals. What the universal domains provide is only the
framework of the structure of linguistic meanings.
In the UNITYP project of Cologne, those cognitive domains that manifest themselves in the
grammars of languages have been analyzed at a cross-linguistic level and assumed to be
universal. The list in F3 is not meant to be complete or definitive in any sense, but may
suffice to give an impression of what is involved.
16
Cf. Langacker 1987 for this conception.
18
Christian Lehmann
F3. Functional domains of the synthetic system
1. Nomination: an entity is named by a descriptive expression or a label (Seiler 1975).
2. Apprehension: an entity is grasped by categorizing and indiviualizing it (Seiler &
Lehmann (eds.) 1982, Seiler & Stachowiak (eds.) 1982, Seiler 1986).
3. Attribution: a representation is modified so that the concept is enriched or the object is
identified (Seiler 1978, Lehmann 1984).
4. Possession: the relation of an entity to another one is represented as inherent in one of
them or established between them (Seiler 1983).
5. Quantification: the extent of the involvement of a set of entities in a predication is
specified.
6. Reference: a representation is determined so that it can be related to and delimited within
the universe of discourse (Seiler 1978).
7. Participation: a situation is articulated into an immaterial center and a set of participants
and circumstants linked to it in various ways.
8. Spatial orientation: an entity is localized in the referential world, this is tied to the
universe of discourse, and this is anchored in deixis.
9. Temporal orientation: a situation is designed with respect to its internal temporal
structure, its temporal limits and relations at various levels.
10. Intensification and comparison: a concept or a thought is assessed qualitatively by
explicit or implicit contrast with similar ones.
11. Nexion: a situation is expanded into a complex one, or several situations are linked
together (Lehmann 1988[T]).
12. Functional sentence perspective: a thought is articulated into subject and predicate (cf.
Sasse 1987, Himmelmann 1988), topic and comment, focus and background.
13. Modality: a thought is rendered relative to illocution and reality.
From the bibliographic references, it is apparent that only some of these have been worked
out in the UNITYP context; the others are projected. A detailed account of the internal
structure of the former may be found in the sources. Very little is as yet known about the set
as a whole, what constitutes membership in it, how the set is structured, whether the domains
differ only by being situated in different realms of cognitive space or also by logical or
semiotic properties, etc.
Each of these domains has an internal structure. Within each of them, a number of techniques
are available which translate the universal concepts and operations in specific languages. A
couple of functional principles account for the ordering of these techniques. For example, in
the domain of attribution, relative clauses, participials and adjectives of various kinds are
ordered on a continuum whose poles are defined by the functions of identification of an entity
vs. enrichment of a concept. By virtue of this gradient internal organization, the functional
domains are the locus of synchronic and diachronic variation.17
17
This is why they are called dimensions in the UNITYP framework.
Language description and general comparative grammar
19
The synthetic system of a GCG - and, consequently, of a specific grammar - is organized
according to these domains and their internal structure. It thereby satisfies all the postulates
P3 - P7.
4.2. The analytic system
The analytic part of the core description is organized independently of linguistic meaning
structures and, instead, based on the structural forms possible in linguistic expressions. Such
an approach has been taken in several schools of European and American structuralism,
including transformational grammar. A number of alternative proposals have been put
forward, but no generally recognized theory of linguistic structure has come forward so far.
The following proposal must be regarded as tentative.
In every language, most of the meanings are expressed by complexes of phonological units
which constitute the significantia of morphemes. These are inventorized in the lexicon. Some
of them are grammatical formatives and therefore reappear in the grammar, but in a different
perspective.
The analytic system is based on the paradigmatic and syntagmatic relations of significantia.
The expression aspects of these are:
paradigmatic:
possibility of substituting a significans in its position by another one or by
zero; relationships of opposition, complementary distribution and free
variation;
syntagmatic: modification of a significans by segmental alternation, intonation or accent;
sequential relationship of a significans to a neighbouring one, including
permutability and bondedness (phonological ties, separability etc.).
However, it would be unwonted and impractical if these structural aspects were the primary
principle of disposition in the grammar. Instead, they are used to define levels of linguistic
structure, units occupying these levels and subclasses of such units. For these, then, the
paradigmatic and syntagmatic relations are specified. This leads to the organization in F4.18
F4. Structural hierarchy of the analytic system
1. Units of different grammatical levels
(Word, syntagm [phrase], clause, sentence, paragraph)
2. [For each unit:] Subclasses of unit
(Word classes, syntactic categories, clause types, sentence types, paragraph types)
3. [For each subclass of unit:] Internal syntagmatic structure, according to the following
types of structural device:
18
F4 follows both Heger 1976 and Coseriu 1987 in treating grammatical levels (ranks) and units as a
primary structuring device. However, in neither of the two models, such concepts would be defined inside
the analytic system.
20
Christian Lehmann
- structural relations between the members of the syntagm [i.e. the units of the next
lower level], in particular: distribution of the members, including positions for
combination of peripheral elements with the head;
- morphological modification [of the flexional type];
- segmental phonological modification;
- prosodic modification.
4. [For each type of structural device:] Paradigmatic relations between members of the
type.
For an example, let us take a particular path through this hierarchy. At the first level, we deal
with the syntagm unit (where `syntagm' is to be taken in the narrow sense of `word group').
At the second level, we subclassify the syntagms and thus come to the noun phrase. At the
third level, we display the internal structure of the noun phrase: we give its constituents,
among them the noun, the adjective attribute, the article; we show the structural positions of
these and the processes of modification which they undergo in the combination. At the fourth
level, we discuss the difference between prenominal and postnominal position of an attribute,
between intonation patterns of the noun phrase, etc.
The hierarchical character of the system in F4 is not equally pronounced at all levels. Certain
grammatical processes (e.g. emphatic accent) may be the same for several of the levels.
However, there are level-specific processes, such as infixation or vowel harmony. These
testify to the essential hierarchical character of linguistic structure.
The fact that terms familiar from English grammar appear in F4 should mislead no one into
thinking that here the structure of English is once again made the basis of a general grammar.
Some languages, such as Dyirbal, do not have a noun phrase, although they may have nouns
and adjectives and syntactic relations between the two. Other languages, such as Turkana and
Guarani, have no or almost no adjectives. While the synthetic system cares for the question
of how the function of the English noun phrase and adjective are fulfilled in these languages,
the grid of the analytic systems of Dyirbal and Turkana will just remain unoccupied at the
places occupied by the noun phrase and the adjective in English. Again, in the analytic
grammars of the former two languages, the structural process of correlative morphological
modification of a noun and its attribute, known as agreement, will play a prominent role, but
not appear in English.
The terms appearing in F4 thus refer to prototypical elements of linguistic structure which
tend to recur, under one form or another, in different languages. Given that expression
devices vary in a gradient fashion, there is a continuum between the concepts of all the levels
of F4. There is no clear boundary between phrase and clause, nor is there one between
sentence types, between types of morphological and phonological modification or between
the various paradigmatic relations. The analytic system of a GCG thus presents a framework
of systematically varying possibilities which assume, in large part, the form of continua. It
provides for the structural possibilities of natural languages in a systematic, but
non-peremptory way and, thus, again satiesfies postulates P3 - P7.
Language description and general comparative grammar
21
5. Conclusion
Throughout the history of general linguistics, two opposing views have been held as regards
the form of a language description. The universalists have maintained that all languages are
fundamentally alike, therefore there must be a general model of language structure which can
be applied in the description of any language. The relativists have argued that every language
is a homogeneous system, therefore a property of one language is never like a property of any
other language, and consequently every language has to be described in its own terms. Both
are right and wrong. There are universals of language, and consequently there can be a
framework for the description of any language. However, there is no universal grammar,
since the association of expression with content is done inside the specific language, not at
a universal level.
The GCG as proposed here tries to solve both the theoretical problem of the above antinomy
and the practical problem of guaranteeing complete and comparable language descriptions.
It achieves this by virtue of two properties: First, it reflects in a systematic way the
cross-linguistic variation and thus provides for the variants appearing in any one language.
Second, like the description of the individual language system, it is organized in a synthetic
and an analytic system. Associations of expression with content which are motivated by
similarity relations of one or the other are therefore captured in a principled way.
Finally, the GCG provides a maximally language-independent and neutral framework which
fits the concepts, categories, relations and processes of any language. It does this by ordering
them on cross-linguistic continua which are the locus for synchronic and diachronic,
interlingual and intralingual, variation.
References
Brettschneider, Gunter & Lehmann, Christian (eds.) 1980, Wege zur Universalienforschung.
Sprachwissenschaftliche Beiträge zum 60. Geburtstag von Hansjakob Seiler. Tübingen:
G. Narr.
Comrie, Bernard & Smith, Norval 1977, "Lingua Descriptive Studies: questionnaire". Lingua
42:1-72.
Coseriu, Eugenio 1987, Formen und Funktionen. Studien zur Grammatik. Tübingen: M.
Niemeyer (Konzepte der Sprach- und Literaturwissenschaft, 33).
Daneš, František 1966, "The relation of centre and periphery as a language universal". TLP
2:9-21.
Dressler, Wolfgang U. 1967, "Wege der Sprachtypologie". Die Sprache 13:1-19.
Edmondson, Jerrold A. & Plank, Frans 1978, "Great expectations: an intensive self analysis".
Linguistics and Philosophy 2:373-413.
Gabelentz, Georg von der 1901, Die Sprachwissenschaft. Ihre Aufgaben, Methoden und
bisherigen Ergebnisse. 2. ed. Reimpr. 1972: Tübingen: G. Narr (TBL, 1).
Greenberg, Joseph H. (ed.) 1963, Universals of language. Report of a conference held at
22
Christian Lehmann
Dobbs Ferry, New York, April 13-15, 1961. Cambridge, Mass.: MIT Press.
Heger, Klaus 1976, Monem, Wort, Satz und Text. Tübingen: Niemeyer (Konzepte der Sprachund Literaturwissenschaft, 8). 2. enlarged ed.
Hewitt, B. George 1979, Abkhaz. Amsterdam: North-Holland (LDS, 2).
Himmelmann, Nikolaus 1987, Morphosyntax und Morphology - Die Ausrichtungsaffixe im
Tagalog. München: W. Fink (Studien zur Theoretischen Linguistik, 8).
Jespersen, Otto 1924, The philosophy of grammar. London: Allen & Unwin.
Langacker, Ronald W. 1987, Foundations of cognitive grammar. I: Theoretical prerequisites.
Stanford: Stanford UP.
Lehmann, Christian 1980, "Aufbau einer Grammatik zwischen Sprachtypologie und
Universalistik". Brettschneider & Lehmann (eds.) 1980:29-37.
Lehmann, Christian 1982, "Some current views of the language universal". LeSt 17:91-111.
Lehmann, Christian 1983, "Directions for interlinear morphemic translations". FoL 16,
1982:193-224.
Lehmann, Christian 1984, Der Relativsatz. Typologie seiner Strukturen, Theorie seiner
Funktionen, Kompendium seiner Grammatik. Tübingen: G. Narr (LUS, 3).
Lehmann, Christian 1987, Theoretical implications of processes of grammaticalization.
Conference on The Role of Theory in Language Description, Ocho Rios, Jamaica, Oct 31
- Nov 8, 1987.
Lehmann, Christian 1988, "Predicate classes and participation". Lehmann, Ch., Studies in
general comparative linguistics. Köln: Institut für Sprachwissenschaft der Universität
(akup, 71); 33-77.
Lehmann, Christian 1988, "Towards a typology of clause linkage". Haiman, John &
Thompson, Sandra A. (eds.), Clause combining in grammar and discourse. Amsterdam:
J. Benjamins; 181-225.
Moravcsik, Edith A. 1972, "Some cross-linguistic generalizations about intensifier
constructions". CLS 8:271-277.
Mosel, Ulrike 1987, Inhalt und Aufbau deskriptiver Grammatiken (How to write a grammar).
Köln: Institut für Sprachwissenschaft der Universität (Arbeitspapier Nr. 4 [N.F.]).
Sapir, Edward 1921, Language. An introduction to the study of speech. New York: Harcourt,
Brace & World (Harvest Books, 7).
Sasse, Hans-Jürgen 1987, "The thetic/categorical distinction revisited". Linguistics
25:511-580.
Seiler, Hansjakob 1969, "On the interrelation between text, translation, and grammar of an
American Indian language". Linguistische Berichte 3:1-17.
Seiler, Hansjakob 1972, "Universals of language". Leuvense Bijdragen 61: 371-393.
Seiler, Hansjakob 1975, "Die Prinzipien der deskriptiven und der etikettierenden
Benennung". Seiler, H. (ed.), Linguistic workshop III. Arbeiten des Kölner
Universalienprojekts 1974. München: W. Fink (Structura, 9); 2-57.
Seiler, Hansjakob 1978, "Determination: A functional dimension for inter-language
comparison". Seiler, H. (ed.), Language universals. Papers from the Conference held at
Gummersbach/Cologne, Germany, October 3-8, 1976. Tübingen: G. Narr (TBL, 111);
Language description and general comparative grammar
23
301-328.
Seiler, Hansjakob 1983, POSSESSION as an operational dimension of language. Tübingen:
G. Narr (LUS, 2).
Seiler, Hansjakob 1986, Apprehension. Language, object, and order. Part III: The universal
dimension of apprehension. Tübingen: G. Narr (LUS 1, III).
Seiler, Hansjakob & Lehmann, Christian (eds.) 1982, Apprehension. Das sprachliche
Erfassen von Gegenständen. Teil I: Bereich und Ordnung der Phänomene. Tübingen: G.
Narr (LUS 1, I).
Seiler, Hansjakob & Stachowiak, Franz Josef 1982, Apprehension. Das sprachliche Erfassen
von Gegenständen. Teil II: Die Techniken und ihr Zusammenhang in den Einzelsprachen.
Tübingen: G. Narr (LUS 1, II).
Talmy, Leonard 1985, "Lexicalization patterns: semantic structure in lexical forms". Shopen,
Timothy (ed.) 1985, Language typology and syntactic description. Cambridge etc.:
Cambridge UP; vol.III:57-149.
Uhlenbeck, Eugenius M. 1980, "Language universals, individual language structure and the
Lingua Descriptive Series Project". Brettschneider & Lehmann (eds.) 1980:59-64.